AI 'Band' Plot Thickens: ‘Spokesperson’ Now Says He’s a Hoaxer

In a bizarre twist that feels like a scene from a digital-age mystery novel, the individual who claimed to be a spokesperson for viral AI band The Velvet Sundown has now revealed himself to be a hoaxer.

The self-proclaimed representative, who identified himself as Andrew Frelon, recently confirmed that the band created their music using the AI tool Suno. However, in a lengthy Medium post published shortly after, Frelon admitted that his entire spokesperson role was an “elaborate hoax” designed to test journalists and media outlets (as per Rolling Stone).

Adding to the confusion, The Velvet Sundown’s official Spotify page has now posted a notice explicitly stating that Frelon has no affiliation with the band whatsoever. This comes as quite a shock considering the band has amassed more than 700,000 listeners on the streaming platform.

In his Medium confession, Frelon claims he capitalised on the band’s minimal social media presence by transforming an account he created in March into what appeared to be the band’s “official account.” He employed various “social engineering” tactics to establish credibility, including strategic interactions with media outlets.

“Knowing from past projects something about the dynamics of journalistic news coverage,” Frelon wrote, “I thought it would be funny to start calling out journalists in a general way about not having reached out to ‘us’ for commentary.”

Interestingly, during his conversation with Rolling Stone, Frelon mentioned his strong interest in “art hoaxes.” The phone number he used to contact the publication now returns a “not in service” message when dialled.

Despite the obvious AI-generated nature of their photos and music, The Velvet Sundown has been featured on numerous Spotify playlists, gaining substantial listenership that prompted widespread media coverage. Glenn McDonald, a former “data alchemist” at Spotify, told Rolling Stone that their prominence on the streaming service exemplifies a shift “away from understandable algorithms with strong grounding in actual human listening and communities” and toward AI-driven systems that “can pick songs for recommendations based on characteristics of their audio.”